Re: Making joins involving ctid work for the benefit of UPSERT

On 2014-07-18 10:53:36 -0700, Peter Geoghegan wrote:
> On Fri, Jul 18, 2014 at 10:46 AM, Andres Freund <andres@2ndquadrant.com> wrote:
> > I think the things that use "wierd" visibility semantics are pretty much
> > all doing that internally (things being EvalPlanQual stuff for
> > INSERT/UPDATE/DELETE and the referential integrity triggers). I don't
> > see sufficient reason why we should break away from that here. Yes,
> > there's stuff that cannot be done when it's done internally, but we can
> > live with those not being possible.
> 
> The whole point of what I was proposing was that those semantics would
> only apply to a special tid scan node. Perhaps I missed something, but
> I'm not sure why you'd consider that I was breaking away from that
> here at all.

I don't see why you'd need such a node at all if we had a fully builtin
UPSERT. The whole stuff with ON CONFLICT SELECT FOR UPDATE and then
UPDATE ... FROM c CONFLICTS is too complicated and exposes stuff that
barely anybody will understand, let alone use correctly in queries they
write themselves.

Greetings,

Andres Freund

-- Andres Freund                       http://www.2ndQuadrant.com/PostgreSQL Development, 24x7 Support, Training &
Services